backlog.md

Feature Requests & Backlog

Tracked from user sessions. Status: done | in-progress | pending | wont-do

How to add new feature requests

Drop plain text at the bottom of this file (below the tables). Claude/GitHub Copilot will read it, convert it into a proper row in the right section, and delete the raw text.

---

Completed ✅

Feature	Notes
Sync preview + conflict resolution flow	Added preview, conflict modal resolution, apply-import, and sync-all continuation path.
Sync mirror failover with active-target switch	Primary failure falls back to mirrors in order; first healthy mirror becomes active target.
Multi-language analytics tracking	Added unique-problem and multi-language aggregation in the analytics dashboard.
Sync regression checks in release pipeline	Added dedicated regression test and enforced it in local release plus GitHub release workflow.
Problem multi-select + bulk CRUD	Added select mode in Solutions with bulk delete, bulk re-tag, and bulk JSON export.
Per-problem topic tag editor	Replaced comma-only tags with add/remove chip editor in the ProblemModal Edit tab.
Platform-specific scraping in handlers only	Moved floating AI LeetCode scraping readers into the LeetCode handler platform adapter.
AI chat in library	AIChatsView ships with saved conversations, context-aware prompts, and attached problems.
New chat from AIChatsView	The library has a dedicated "+ New Chat" flow to start conversations from the chat view.
Problem attachment in chats	Multiple problems can be attached to one chat and reused in the conversation context.
Math rendering parity	AI markdown rendering supports inline and block math through the KaTeX renderer.
Mermaid rendering parity	AI markdown rendering supports Mermaid fences with inline render and fallback live-link output.
LeetCode profile import overhaul	GraphQL queries fixed, tags and difficulty fetched, solve timestamps recorded rather than guessed.
Codeforces profile import	user.status back-catalogue import on your own profile. Arrives without code — CF does not publish submission source.
NeetCode and takeuforward handlers	Both ship with their own judges tapped at the network layer; TUF sheets are marked, never committed.

---

In Progress 🔄

Feature	Status	Notes
Multi-platform analytics integration	partial	Graph has platform colors; analytics is still mostly LeetCode-centric.
LeetCode import completeness	partial	Recent 20 via public API; full history still requires the profile page button.
Solutions advanced search & filters	active	Add dropdowns, tag/topic/language filters, and free-text search across title, tags, and overview.
Documentation revamp and dedupe	done	Canonical docs index, archive layout, and root markdown cleanup are complete.

---

Bugs to Fix 🐛

Bug	Priority	Notes
(none currently open)	—	—

Pending 📋

Core UI

Feature	Priority	Notes
ProblemModal enhancements	medium	Show accept rate, hints, and similar problems with links; remove low-value fields.
Custom scrollbar design	medium	Replace stock OS scrollbars with a styled thin scrollbar across the library UI.
LeetCode /progress page stats integration	low	Reference leetcode.com/progress for streak calendar, topic breakdown, and badge data.
Incognito mode timer + "indefinitely" option	medium	Add time selector on incognito toggle with 1h, 4h, 24h, and indefinite options.
Modal persistence	high	Keep modal state across tab switches and expanded sections.

AI Context and Commands

Feature	Priority	Notes
AI command palette on / input	high	Dropdown of commands such as /mycode, /problem, /test, /optimize, and /explain with auto-attached context when relevant.
/mycode - Personal code context	high	Automatically surface the user’s code when it improves the answer.
/problem - Problem context	high	Automatically attach current problem details when relevant.
/test - Test case extractor	high	Generate tests and run them against the editor code before submit.
/optimize - Optimization request	high	Suggest code optimizations with before/after trade-offs.
/explain - Detailed explanation	high	Break down algorithms step by step with clear data-structure reasoning.
/mermaid - Diagram generation	high	Generate Mermaid diagrams inline when they make the explanation clearer.
/math - Math helper	medium	Surface formulas and render math when the response needs it.
/similar-patterns - Pattern matcher	medium	Show similar problems solved and the common technique used.
/complexity - Complexity analyzer	medium	Provide time and space complexity analysis with the reasoning.
Recently used commands	medium	Pin favorite commands to the top and sort by usage frequency.
AI request templates	medium	Save common requests as templates with a quick-insert shortcut.

AI Enhanced Storage

Feature	Priority	Notes
AI chat metadata enrichment	high	Local-only storage for request templates, favorite commands, and per-conversation tags.
Chat response ratings	high	Local feedback for the user’s own history and future prompt selection.
AI session analytics	medium	Track local command usage and response usefulness for personal insights only.

Graph Advanced Visualization 📊

Feature	Priority	Notes
AI submission context (test vs submit)	high	Detect test-case failures versus accepted/rejected submissions and fetch the right analysis.
AI system on LeetCode/platform pages	medium	Floating panel with question plus editor context that calls the configured AI handler.
Enhanced recommendation system	low	Move beyond Blind 75 with personalized recommendations from weak topics and recent solves.
Full multi-platform analytics	low	Bring GFG and Codeforces into heatmap, difficulty chart, and topic breakdown views.
Submission auto-detect investigation	investigate	LeetCode handler MutationObserver may miss accepted results in some layouts.
Graph community detection	high	Detect topic clusters, highlight communities, and suggest learning paths.
Node density filtering	high	Zoom-based density levels: far = topics only, medium = topics plus solved, close = all.
Zoom to node whose modal is open	high	Auto-center the graph on the open problem modal when that improves context.
Multi-node selection	low	Ctrl+Click to select multiple nodes and compare interactions between problems.
Edge label rendering	medium	Show edge types such as topic, similar, and canonical with a visibility toggle.

AI-MCP Integration 🧠

Feature	Priority	Notes
MCP Tool: Query Problems	high	Search by topic, difficulty, platform, or pass rate and return recommendations.
MCP Tool: Get Problem Stats	high	Return solve time, pass rate, difficulty consensus, and acceptance rate.
MCP Tool: Next Problem Suggestion	high	Analyze weak topics and suggest the next best problem with rationale.
MCP Tool: Code Quality Analysis	medium	Run code analysis for complexity, edge cases, and improvement suggestions.
MCP Tool: Trend Analysis	medium	Show learning trends, difficulty progression, and platform distribution.
MCP Tool: Similar Solution Discovery	medium	Find similar problems and extract the pattern or technique used.
MCP Context: User Profile	high	Provide an IndexedDB snapshot of solved problems, weak topics, and preferred platforms.

---

Deferred audit findings 🔍

From the 2026-08 full-source audit. Each was judged real but consciously deferred — the fix is either feature-sized, needs a product decision, or carries more regression risk than the bug it removes. IDs refer to the audit report.

ID	Finding	Why deferred
F10 / M2	index.json is last-writer-wins across devices; the 422 retry rebuilds from a stale head	Needs a union-merge of remote index entries by commit key before every index write — a cross-device merge design of its own, not a spot fix.
F13	init() races onInstalled, and some listeners register after awaits	Restructuring SW listener registration risks breaking event delivery in ways only on-device MV3 testing can catch; initPromise already dedupes most of it.
F21	presence-marker.js writes auth.tokens via two racing paths	The window is one OAuth callback racing itself; collapsing to the SW-relay path needs live OAuth testing to verify no token is dropped.
F23	manuallyEdited overwrite guard only applies under skipCommit	saveProblem's merge semantics already preserve most manual edits; the residual case needs a field-level merge policy decision.
F26	Pervasive silent catch {} around commit-adjacent operations	The worst consequence (invisible errors) was fixed at the root — console.error is never gated now. Auditing every catch individually is its own pass.
F27 / L7	271 direct chrome.* uses outside browser-compat.js	Wholesale mechanical refactor with real regression surface and zero user-visible gain; do it file-by-file as files are touched.
L3	GFG detection includes a body-wide "problem solved successfully" text search	Weakest signal in the codebase but guarded by dedup keys; improving it needs live-DOM selector work on GFG.
L5	TUF slug fallback can attribute a solve to the currently open page	The TTL-guarded pairing already bounds this; fully closing it means carrying the slug through the judge poll.
L11	pages-template interpolates owner/repo into a JS string raw	Exploit requires GitHub's own owner/repo charset validation to break first.

Resolved since the audit (2026-08): F16 — the metadata-refresh tab queue now lives in storage.session and resumeRefreshQueue() picks it back up on every SW wake; F17 — the docs now state the true invariant (index.json in every commit, order irrelevant to the Trees API); F24 — the canonical-map TTL clock is persisted and the 304 path populates the in-memory map from cache; F25 — importChatsLocal dedups by _githubPath inside its own transaction; L1 — an unreadable language now commits as unknown/.txt instead of a guessed "C++"; L4 — event-driven handlers queue a second accepted event behind the in-flight one instead of dropping it; L12 — the manifest domain list is derived from the DOMAINS exports in dom-selectors.js.

---

Won't Do / Deferred ❌

Feature	Reason
Full history via Settings import (unlimited)	LeetCode public API caps at 20; full import must use the profile page button or CLI importer.
Real-time leaderboard	Requires server-side infrastructure and is out of scope for a client-only extension.